Purpose: The purpose of this study was to prove directly whether cells from the blood stream contribute to endothelialization of isolated, impervious Dacron vascular grafts in the dog. Methods: We designed an 18 cm, three-component graft with two parallel central Dacron limbs; one was made impervious with silicone rubber, and the other was preclotted. This model was implanted in the canine descending thoracic aorta with 30 μm polytetrafluoroethylene grafts anastomosed at each end. An 8 cm, three-component graft completely coated with silicone rubber was implanted in the canine abdominal aorta and inferior vena cava. Implant periods ranged from 4 to 12 weeks. Flow surfaces were studied by use of stereomicroscopy after being stained with silver nitrate, and by use of scanning and transmission electron microscopy, the inner wall and flow surface were studied by light microscopy after hematoxylin-eosin and immunocytochemical staining (the latter for endothelial and smooth muscle cells), and the full wall was studied by light microscopy after hematoxylin-eosin staining. Results: Effective prevention of pannus and transmural ingrowth into the impervious central test grafts was achieved, and scattered islands of endothelial cells were conclusively demonstrated on flow surfaces in each of the three implant sites 4 weeks after implantation. In the descending thoracic aorta, where these grafts were also implanted for 8 and 12 weeks, α-actin – positive cells and microvessels were found beneath some of the endothelial islands. Conclusion: Fallout endothelialization of Dacron vascular grafts occurs in both the arterial and venous systems of the dog. (J VASC SURG 1994;20:546-57.)